Abstract: Fourier transform infrared spectroscopy (FTIR) combined with soft independent modeling of class analogy (SIMCA) method was employed to the identification of different varieties of milk. The optimized PCA model was built by leave-one-out cross-validation (LOOCV) method after series of pre-treatments such as baseline correction and Savitzky-Golay smoothing in the region of 3100 — 850 cm-1. Under the α ＝5% significance level, the identification rates of this model for pure milk, low lactose milk, low fat milk and high protein milk were 80%, 80%, 100% and 80%, respectively, and the rejection rates were 93%, 100%, 100% and 93%, respectively. This indicates that FTIR combined with SIMCA is a valid method for rapid identification of different varieties of milk.

Key words: FTIR；SIMCA；milk；pattern recognition

摘要： 利用傅里叶变换红外光谱法（FTIR）结合软独立模式分类法（SIMCA）对不同类别的牛奶进行识别。通过对光谱数据基线校正和Savitzky-Golay平滑处理后,在3100～850cm-1光谱区域,利用留一交互验证法建立获得主成分分析（PCA）最优模型。在α=5%显著水平下,最优模型对纯牛奶、低乳糖奶、低脂奶和高蛋白奶的识别率分别为80%、80%、100%和80%,拒绝率分别为93%、100%、100%和93%。表明FTIR结合SIMCA可成为快速识别牛奶类别的有效方法。

关键词: 傅里叶变换红外光谱法; 软独立模式分类法; 牛奶; 模式识别